To the Editor: Kristin M. von Ranson, Ph.D., and colleagues (1) found higher obsessive-compulsive disorder (OCD) scores in patients who had recovered from bulimia than in a normal comparison group. They suggested the possibility that these behaviors are trait related, since they did not change dramatically with treatment. They neglected to consider, however, an alternative explanation: "the rollback phenomenon." As an illness remits, it progressively recapitulates (although in reverse order) many of the stages and symptoms that were seen during the time it developed (2). The rollback phenomenon has been substantiated in mood disorders (3) and anxiety disorders (4). There is some preliminary evidence that may introduce this possibility also in bulimia nervosa.

A standardized semistructured research interview for eliciting prodromal symptoms (3, 4) was administered to 30 recovered patients with bulimia nervosa (5). Most of the patients reported prodromal symptoms concerned with anxiety, depression, and irritability. Five of the 30 patients reported obsessive-compulsive symptoms according to a specified threshold. The study group included only patients who did not have comorbid axis I DSM-IV disorders. It is conceivable that the number of patients with obsessive-compulsive symptoms could have been higher in a more heterogeneous group. We postulated that the combination of subclinical affective symptoms and stressful life events may create an allostatic load prompting precipitation of bulimia nervosa. Risk factors, such as repeated exposure to negative comments about shape, weight, and eating, may further increase the vulnerability to this allostatic load, resulting in a chronic dissatisfaction with body image and an obsessive preoccupation with food (5). Prodromal symptoms may persist as residual symptoms while the most typical symptom profile abates (3, 4).

The state-trait dichotomy is largely inadequate in clinical settings: psychological constructs conceived as trait dimensions may display sensitivity to change in a specific clinical situation, whereas constructs viewed as state dimensions—because of the rollback phenomenon—may display unexpected stability throughout the longitudinal development of the disorder (6).

Unlike characterological traits, residual symptoms are amenable to treatment, particularly according to a sequential model. This may result in therapeutic efforts of more enduring quality than current strategies (3, 4).